National Library of Scotland says aye to Scality object storage

The National Library of Scotland has deployed Scality Ring object storage across two datacentres, in a move that will see it replace tape as a long-term store of digitised assets. The deployment of object storage – also with file access – will also see the museum replace several existing Hitachi SANs as application storage.

Currently the organisation holds about 1.5PB of digitised assets, including content from books and newspapers, sound recordings from a variety of media right back to wax cylinders, and movies from VHS tape.

The current scheme has digitised data held on seven Hitachi SANs, some of which are nearing end of life, with backups to Qualstar Tape libraries across two datacentres in Edinburgh and Glasgow.

The existing infrastructure was becoming unwieldy, with the very idea of backups becoming unsuited to the workload, said associate director Stuart Lewis.

“Backup doesn’t scale, and is not well suited to data that doesn’t change very much,” he said. “As backups grow in size, it places more of a burden in terms of checking tapes and being able to recover them, and potentially rewriting them as LTO generations arrive.”

Also, the existing seven Hitachi SANs presented a further layer of complexity with dozens of network drives that needed mapping and maintaining. There is also potential contention between digital asset workloads and the library’s VMware virtualised environment.

“So, we started to look for a system that could grow with us,” said Lewis. “And to increase resilience with three copies of everything, rather than the two we have now.”

The National Library eventually plumped for object storage, and has deployed two separate Scality Rings at its two datacentres. In terms of hardware, it is deployed on six HP Apollo 4510 servers with 18 10TB in SAS drives drives and a 960GB NVMe card for metadata on each.

Connectivity to the Scality hardware is via the S3 object storage protocol or the recently deployed CIFS file access connector. Between datacentres, there is a 10GBps fibre connection. Two copies are kept on Scality and a further copy staged off to Amazon Glacier cloud storage.

Scality is object storage, which is very well suited to large amounts of unstructured data, as is likely to be found in digital archives.

It stores objects with unique identifiers in a flat structure, as opposed to network-attached storage (NAS), which keeps data in a tree-like file system structure and can run into performance issues when the data stored runs into very large amounts of files.

So what have been the benefits so far?

“It’s too early to quantify them,” said Lewis. “But when we decommission the SANs we no longer need, we will be able to say we saved x amount of money, as well as the time it has taken staff to administer the seven SANs.”

In terms of space, the contrast is between 100TB per rack with the older Hitachi SANs and up to 4PB now.

The plan is to reduce SAN capacity down to two nodes and eventually replace those with all flash, said Lewis. That way, virtualised workloads and everyday I/O will be split from the digitised assets on Scality, although some file shares will migrate to object storage via the CIFS connector.

Did Lewis have any concerns about moving to object storage, especially as it is by no means a fully mainstream storage technology yet?

“We had to rewrite some local applications for S3, but these had been developed in-house, so didn’t need to rely on vendors for that,” he said. “And S3 is now very mature, with lots of code libraries available.

“Also, we wanted resilience via the use of multiple technologies. We’ve got that by being able to send data to Amazon, and that is via S3 too.”